Endogenous dendritic cells from the support T-ALL growth via IGF1R activation

Todd A. Tripletta, Kim T. Cardenasa,1, Jessica N. Lancastera, Zicheng Hua, Hilary J. Seldena, Guadalupe J. Jassoa, Sadhana Balasubramanyama, Kathy Chana, LiQi Lib, Xi Chenc,d, Andrea N. Marcogliesee, Utpal P. Davéf, Paul E. Loveb, and Lauren I. R. Ehrlicha,2

aDepartment of Molecular Biosciences, Institute for Cellular and Molecular Biology, The University of Texas at Austin, Austin, TX 78712; bSection on Hematopoiesis and Lymphocyte Biology, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD 20892; cDivision of Biostatistics, Department of Health Sciences, University of Miami Miller School of Medicine, Miami, FL 33136; dSylvester Comprehensive Center, University of Miami Miller School of Medicine, Miami, FL 33136; eDepartment of Pathology and Immunology, Baylor College of Medicine, Houston, TX 77030; and fDivision of Hematology/Oncology, Tennessee Valley Healthcare System and Vanderbilt University Medical Center, Nashville, TN 37232

Edited by Zena Werb, University of California, San Francisco, CA, and approved January 14, 2016 (received for review October 15, 2015) Primary T-cell acute lymphoblastic (T-ALL) cells require tumor growth and metastasis (3). Tumor-associated stromal-derived signals to survive. Although many studies have (TAMs), which resemble alternatively activated (M2) macro- identified cell-intrinsic alterations in signaling pathways that promote phages (4), also support tumor growth. TAMs suppress antitumor T-ALL growth, the identity of endogenous stromal cells and their immune responses, promote tumor invasion and , and associated signals in the tumor microenvironment that support T-ALL are negatively associated with clinical outcomes (5). remains unknown. By examining the thymic tumor microenviron- depletion slows tumor progression in multiple cancer models (6–8). ments in multiple murine T-ALL models and primary patient samples, Dendritic cells (DCs) within tumor microenvironments also have we discovered the emergence of prominent epithelial-free regions, aberrant phenotypes and have been implicated in suppression of enriched for proliferating tumor cells and dendritic cells (DCs). antitumor adaptive immune responses (9). Thus, through mitogenic Systematic evaluation of the functional capacity of tumor-associated signaling, tissue remodeling, and suppression of adaptive immunity, stromal cells revealed that myeloid cells, primarily DCs, are necessary heterogeneous cells in the tumor microenvironment cooperate and sufficient to support T-ALL survival ex vivo. DCs support T-ALL to promote growth of solid and hematologic (2). growth both in primary thymic tumors and at secondary tumor sites. Importantly, human primary T-ALL samples have been shown to To identify a molecular mechanism by which DCs support T-ALL require stromal-derived signals for survival (10). Thus, identifica- growth, we first performed expression profiling, which revealed up-regulation of platelet-derived growth factor beta (Pdgfrb) tion of tumor types that support T-ALL could unveil and -like growth factor I receptor (Igf1r)onT-ALLcells,with novel therapeutic targets. concomitant expression of their ligands by tumor-associated DCs. There is ample reason to suspect that the thymic microenviron- Both Pdgfrb and Igf1r were activated in ex vivo T-ALL cells, and co- ment, the primary site of T-ALL, contributes to lymphomagenesis. culture with tumor-associated, but not normal thymic DCs, sustained IGF1R activation. Furthermore, IGF1R signaling was necessary for DC- Significance mediated T-ALL survival. Collectively, these studies provide the first evidence that endogenous tumor-associated DCs supply signals driv- T-cell acute lymphoblastic leukemia (T-ALL) is a of ing T-ALL growth, and implicate tumor-associated DCs and their mi- developing T cells. Cancer cell growth is often driven by cell- togenic signals as auspicious therapeutic targets. intrinsic alterations in signaling pathways as well as extrinsic signals from the tumor microenvironment. Here we identify T-ALL | leukemia | dendritic cell | tumor microenvironment | IGF1R tumor-associated dendritic cells as a key endogenous cell type in the tumor microenvironment that promotes murine T-ALL -cell acute lymphoblastic leukemia (T-ALL) is an aggressive growth and survival at both primary and metastatic tumor sites. Tmalignancy of T-cell progenitors that predominantly affects We also find that tumor-associated dendritic cells activate the in- children and adolescents. Intensified chemotherapeutic regimens sulin-like growth factor I receptor in T-ALL cells, which is critical for have improved 5-y event-free survival rates to over 75% in children their survival. Analysis of primary patient T-ALL samples reveals and 50% in adults. However, these therapies are toxic, and patients phenotypically analogous tumor microenvironments. Our findings who fail to respond or relapse have poor outcomes (1). The main suggest that targeting signals from the tumor microenvironment focus of T-ALL research has been identification of cell-intrinsic could expand therapeutic options for T-ALL. genetic permutations that promote tumor growth, such as activating in the transcription factor NOTCH1,deletionoftumor Author contributions: T.A.T., K.T.C., and L.I.R.E. designed research; T.A.T., K.T.C., J.N.L., suppressor in the CDKN2A , and aberrant activation of H.J.S., G.J.J., S.B., and K.C. performed research; L.L., A.N.M., U.P.D., and P.E.L. contributed new reagents/analytic tools; T.A.T., K.T.C., J.N.L., Z.H., K.C., X.C., and L.I.R.E. analyzed transcriptional regulators such as LIM domain only 2 (LMO2)(1). data; and T.A.T. and L.I.R.E. wrote the paper. Despite our understanding of these common cell-intrinsic alterations, The authors declare no conflict of interest. we have yet to develop successful, targeted therapeutic approaches to This article is a PNAS Direct Submission. improve patient outcomes, suggesting that we need to broaden our Data deposition: The transcriptional profiling data reported in this paper have been de- understanding of other factors promoting T-ALL growth (1). posited in the Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo Tumor progression is driven not only by cell-intrinsic alter- (accession no. GSE54609). ations but also by dynamic interactions between cancer cells 1Present address: Department of Molecular Medicine, University of Texas Health Science and the surrounding tumor microenvironment (2). For example, Center at San Antonio, San Antonio, TX 78240. cancer-associated (CAFs) secrete factors that pro- 2To whom correspondence should be addressed. Email: [email protected]. mote tumor survival, angiogenesis, and recruitment of inflamma- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. tory cells. CAFs also mediate tissue remodeling that promotes 1073/pnas.1520245113/-/DCSupplemental.

E1016–E1025 | PNAS | Published online February 9, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1520245113 Downloaded by guest on September 27, 2021 Normal thymocytes must engage in bidirectional signaling with het- PNAS PLUS erogeneous thymic stromal cells to promote proper differentiation of both thymocyte and stromal compartments (11, 12). Thymic stromal cells consist of hematopoietic cells, such as DCs and macrophages, and nonhematopoietic cells, such as thymic epithelial cells (TECs) and fibroblasts. Thymocyte:TEC crosstalk is essential at multiple stages of T-cell development. TECs express NOTCH ligands and IL-7, which are required for differentiation of normal T-cells (13–15), buthavealsobeenimplicatedinT-ALLtumorigenesis(10,16).Thus, nascent cells may coopt signals in the thymic microenvi- ronment to promote tumor growth. Previous studies demonstrated that stromal support is required for ex vivo survival of mouse and human T-ALL cells (10, 17–20) and implicated IL-7 and NOTCH1 signaling in promoting tumor survival (10, 16–18). Together, these findings suggest that TECs might promote T-ALL. However, the above studies addressed the role of stroma in supporting T-ALL predominantly using cell lines and exogenous cytokines; thus, the contribution of endogenous stromal cells in the T-ALL tumor mi- croenvironment remains largely unexplored. Herein, we identify primary stromal cells from the endogenous tumor microenvironment that support T-ALL growth ex vivo. Our data reveal that DCs are expanded within primary and metastatic Fig. 1. LN3 thymic T-ALL cells require tumor stroma to survive and pro- liferate ex vivo. (A) Survival of indicated strains shown as Kaplan–Meier T-ALL tumor microenvironments. Despite the essential contribution survival plots (n = 33, WT; n = 155, LN3). (B–E) Primary T-ALL cells were of TECs to normal thymocyte development, TECs are dispensable cultured with or without stromal cells enriched from mechanically dissoci- for T-ALL survival in tumor:stromal cocultures. In contrast, myeloid ated LN3 thymic tumors or WT thymi. Representative flow cytometric plots cells, particularly DCs from tumor microenvironments at both pri- evaluating viability (B) or DNA content (D) of viable cells 6 d after culture mary and secondary tumor sites, are necessary and sufficient to initiation. (C and E) The percentage of viable lymphoma cells (C) and viable support growth of T-ALL cells. Tumor-associated, but not wild-type cells in S/G2/M (E) at the indicated number of days in culture. Graphs depict (WT), thymic DCs support T-ALL survival and activate IGF1R and means + SD of cumulative data from duplicate wells from four independent < < downstream mitogen-activated protein kinase (MAPK) signaling in experiments. *P 0.05 and **P 0.01. T-ALL cells. Importantly, IGF1R activation is required for DC- mediated T-ALL survival. Analysis of primary T-ALL patient tissues The Thymic Epithelial Compartment Is Progressively Disrupted During reveals similar tumor microenvironments, consisting of expansive T-ALL Tumorigenesis. Because tumor-associated, but not WT thymic, epithelial-free regions with infiltrating DCs. Together, our findings stromal cells supported lymphoma growth, we investigated com- reveal an unanticipated role for endogenous tumor-associated positional changes in the thymic microenvironments of LN3 mice DCs in directly promoting survival of T-ALL cells through an during T-ALL progression. We first evaluated TECs because they IGF1R-dependent mechanism, suggesting tumor-associated DCs provide signals essential for thymocyte development, including IL-7 or their corresponding mitogenic signals might serve as auspicious and NOTCH ligands. In WT mice, the thymic epithelium consists of therapeutic targets. + central Keratin-5 medullary TECs (mTECs) surrounded by + Results Keratin-8 cortical TECs (cTECs). In LN3 mice, TEC organization was abnormal as early as 1 mo of age, with enlarged mTEC regions Primary T-ALL Cells Require Tumor Stroma for Survival ex Vivo. To investigate the contribution of tumor stroma to T-ALL growth, we extending to the thymic capsule by 3 mo (Fig. 2A). In comparison used the LN3 murine model of T-ALL in which endogenous TCRαβ with small isolated epithelial-free regions that can be found in the rearrangements promote spontaneous lymphomagenesis (21). Al- interior of some WT thymi (Fig. 2A) as previously described (23), though the oncogenic lesions driving T-ALL in LN3 mice have not strikingly large peripheral epithelial-free regions were abundant in been elucidated, activation of NOTCH1 and NFAT were previously LN3 thymic , as seen in higher-magnification images implicated (21). Notch1 exon sequencing revealed nonsynonymous (Fig. 2A) and in whole-thymic sections (Fig. 2B). These epithelial- mutations in the heterodimerization or PEST domains in 64% (9/14) free regions first appeared at the perimeter of pretumor LN3 thymi of primary LN3 tumors, with premature STOP codons in the PEST at 1 and 3 mo of age (Fig. 2 A and B). Thus, the thymic epithelium domains of 42% (6/14) of LN3 T-ALLs (Table S1). Thus, T-ALL in is progressively disorganized, culminating in expansion of extensive the LN3 model shares characteristics with human NOTCH-driven epithelial-free regions during LN3 T-ALL development. cortical T-ALL, including constitutive NOTCH signaling and NFAT activation (Table S1) (19, 21, 22), as well as expression of CD8 with Thymic Epithelial-Free Regions Are Enriched for Proliferating Lymphoma variable levels of CD4 (21), and elevated expression of the NOTCH1 Cells Surrounded by Vasculature, Fibroblasts, and DCs. TECs are es- transcriptional targets Hes1, Dtx1,andpTCRa (https://gexc.stanford. sential for thymocyte development and can support T-ALL survival in vitro (18), which led us to question whether tumor proliferation edu/models/1118/genes). We found that 60% of LN3 mice developed + T-ALLby1yofage(Fig.1A). could be supported in epithelial-free regions. Surprisingly, CD8 To determine whether endogenous tumor stromal cells support tumor cells proliferate extensively in these regions, suggesting that T-ALL growth, we cultured primary LN3 T-ALL cells in the pres- nonepithelial cells in the tumor microenvironment likely support MEDICAL SCIENCES ence or absence of stroma from WT or tumor-bearing thymi. T-ALL growth (Fig. 3A and Fig. S1). To identify such alternate cell Whereas tumor cells cultured alone or with WT thymic stroma types, we quantified stromal cells from WT versus tumor-bearing underwent cell death, T-ALL cells cultured with tumor-associated LN3 thymi (phenotypes and gating strategies in Fig. S2). cTECs stroma survived (Fig. 1 B and C). Furthermore, a significantly were largely absent in LN3 tumors, whereas mTECs were ex- + higher proportion of viable T-ALL cells proliferated in the pres- panded. Strikingly, both the number and percentage of Sirpα DCs − + + + + ence of tumor-associated stroma (Fig. 1 D and E). These findings (CD11chiEpCAM CD45 MHCII Sirpα CD80 cells), one of two indicate that stromal cells in the thymic microenvironment become conventional thymic DC subsets (24), were greatly increased in altered to support T-ALL survival and proliferation. tumor-bearing thymi (Fig. 3 B and C). Immunostaining revealed

Triplett et al. PNAS | Published online February 9, 2016 | E1017 Downloaded by guest on September 27, 2021 T-ALL Cells Require Contact with Tumor Stroma for Survival and Interact Extensively with DCs in the Thymic Microenvironment. Transwell assays revealed that T-ALL cells required close con- tact with tumor stroma to survive (Fig. 4A), similar to reports for chronic lymphocytic leukemia (26). Thus, we determined whether T-ALL cells made preferential contacts with stromal subsets in intact 3D thymic microenvironments. Using our pre- viously established two-photon imaging approach (27), we compared the migration of WT thymocytes with LN3 tumor cells + in live thymic slices containing EYFP DCs (Fig. 4B) (28). No- tably, LN3 T-ALL cells were in contact with thymic DCs twice as often as WT thymocytes (Fig. 4C) and exhibited a 190% increase in dwell time upon DC contact (Fig. 4D). Thus, LN3 tumor cells require contact with tumor stroma for survival ex vivo and in- teract extensively with thymic DCs in situ, which are positioned to provide mitogenic signals in epithelial-free regions within the tumor microenvironment.

Fig. 2. Thymic epithelial-free regions emerge during T-ALL development. (A) Thymic cryosections from 1-, 3-, and 6-mo-old WT mice, from 1- and 3-mo- old tumor-free LN3 mice, and from a 6-mo-old tumor-bearing LN3 mouse, were immunostained for Keratin-5 (K5) and Keratin-8 (K8) with or without DAPI over- lays. (Scale bar: 200 μm.) Borders of epithelial-free regions are denoted by white dotted lines. (B) Entire thymic cryosections from WT and LN3 mice immunostained as in A reveal overall changes in organization of the epithelium and emergence of peripheral epithelial-free regions as LN3 tumors form. (Scale bars: 1 mm.) Images in A and B are representative of three thymi at each stage and genotype.

+ that Sirpα DCs were present in peripheral epithelial-free regions of LN3 thymic lymphomas; in contrast, DCs localize within central medullary regions of WT thymi (Fig. 3D). Epithelial-free regions in LN3 thymic tumors were also highly vascularized and contained extensive / (ECM) networks (Fig. 3E), consistent with a microenvironment supportive of tumor growth. Similar epithelial-free regions containing DCs, vasculature, and fi- broblast networks were observed in thymic lymphomas from LMO2 transgenic mice, which overexpress an frequently up- regulatedinhumanT-ALL(25)(Fig.3D and E). Primary patient Fig. 3. Proliferating tumor cells are enriched in thymic epithelial-free regions T-ALL mediastinal mass sections also consisted of extensive epi- containing DCs, fibroblasts, and vasculature. (A) Pan-Cytokeratin (Pan-CK) and thelial-free regions containing dense DC networks, in contrast to BrdU immunostaining following a 20-min BrdU pulse reveals extensive pro- the encompassing epithelial network and central DC localization in liferation of T-ALL cells in thymic epithelial-free regions. (Scale bar: 200 μm.) − F (B and C) Cellularity of thymic stromal subsets (B) and percentages of Sirpα and normal human thymi (Fig. 3 ). All thymic lymphomas analyzed + from the two mouse models and patient samples contained exten- Sirpα DCs (C) in LN3 thymic lymphomas versus WT thymi from mice 6 mo of = = sive epithelial-free regions with DC networks, suggesting com- age. Graphs depict cumulative data from multiple experiments (n 6, WT; n 4, LN3 tumors); bars represent means. (D and E) WT and tumor-bearing LN3 and monality between tumor microenvironments in murine T-ALL of LMO2 thymic cryosections immunostained for Pan-Cytokeratin (Pan-CK), CD11c, diverse oncogenic origins and in human T-ALL. and Sirpα (D) or Pan-CK, ER-TR7, and MECA-32 (E). (Scale bars: D, Left and E, As in overt tumors, early epithelial-free thymic regions in young 200 μm; D, Center and Right, 100 μm.) (F) Representative images of normal asymptomatic LN3 mice contained proliferating thymocytes (Fig. human thymic and T-ALL patient mediastinal mass sections immunostained for + S3A), Sirpα DCs (Fig. S3B), vasculature, and fibroblasts (Fig. S3C). Pan-CK and CD11c. (Scale bars: Left, 100 μm; Center and Right,50μm.) Immu- + = = = In addition, the number and percentage of Sirpα DCs increased nostains are representative of n 4(A), n 3(D and E), and n 4(F) biological – replicates each. White dashed lines outline medullary regions (M) (thin lines) in with age in pretumor LN3 thymi but not in WT thymi (Fig. S3 D G). WT/normal thymi and epithelial-free regions (thick lines) in thymic lymphomas. Thus, microenvironments comparable to those in overt lymphomas Few epithelial cells were identified in patient mediastinal mass sections. *P < emerge early and expand during T-ALL progression. 0.05, **P < 0.01, and ***P < 0.001.

E1018 | www.pnas.org/cgi/doi/10.1073/pnas.1520245113 Triplett et al. Downloaded by guest on September 27, 2021 Fig. 4. LN3 lymphoma cells require close contact PNAS PLUS with stroma for ex vivo survival and make frequent and prolonged contacts with DCs in an intact thymic microenvironment. (A) Cocultures were initiated us- ing transwell inserts to compartmentalize lymphoma cells and thymic tumor stromal cells, as indicated. Red axis labels indicate the chamber analyzed for viability after 6 d. Data from three experiments are shown, with each symbol representing the mean of duplicate wells. (B–D) Labeled WT thymocytes or LN3 lymphoma cells were imaged by two-photon microscopy in thymic slices from CD11c-EYFP mice. (B) Trajectories of WT and LN3 tumor thymocytes (red) migrating among thymic DCs (yellow) are displayed as color-coded tracks from start (blue) to end (red) of the time course. White spheres indicate tracked cells. See corresponding movies (Movies S1 and S2). Cell tracking was used to calculate the percentage of time points in which cells contacted DCs (C) and the duration of DC contacts (D). Graphs represent cumulative data from five experiments. (C) Data were calculated from cumulative time points (n = 15,706: WT; n = 13,048: LN3) of all tracked cells. (D) Means ± SD from cumulative tracked cells (n = 505, WT; n = 306, LN3) are displayed in red on the plots. Dots represent individual cells; mean values are displayed above the graph. *P < 0.05 and ****P < 0.0001.

Myeloid Cells from Primary Tumor Microenvironments Promote ence of tumor-associated stroma from their respective organs Lymphoma Survival ex Vivo. To identify the endogenous tumor (Fig. 6A). Flow cytometric analyses revealed a significant in- + − stromal cell types capable of supporting survival of LN3 T-ALL crease in the number of DCs (CD11c B220 ) and a consistent, cells, we cocultured primary tumor cells with sorted stromal subsets although nonstatistically significant, increase in the number of − + from the same tumors. Cocultures containing a combination of all CD11c CD11b myeloid cells in tumor-bearing lymph nodes sorted stromal subsets (“All Stroma”)resultedinanaverageof52% and , resulting in maintenance of the DCs’ proportional lymphoma viability after 6 d (range of viability, 24–78%; stromal representation in these metastatic organs (Fig. 6B and Fig. S5 phenotypes and sort purities in Fig. S4 A–C). To control for variable B and C). Importantly, tumor-associated splenic DCs were nec- survival between experiments with different primary tumors, we essary and sufficient to promote survival of splenic T-ALL cells normalized T-ALL viability to cultures containing All Stroma for (Fig. 6C) and could also support survival of T-ALL cells from each experiment. Depletion of epithelial cells or fibroblasts had no primary thymic lymphomas and metastatic lymph nodes (Fig. effect on tumor survival ex vivo (Fig. 5A). Strikingly, depletion of 6D). Conversely, splenic versus thymic DCs from the same tu- myeloid cells or DCs alone resulted in a significant decrease in the mor-bearing mice had a comparable capacity to support survival viability of T-ALL cells (Fig. 5A). Conversely, sorted tumor-associ- of splenic T-ALL cells (Fig. 6E). Together, these data demon- − + atedDCsand/orCD11cCD11b myeloid cells were sufficient to strate that tumor cells from primary and metastatic tumor sites support T-ALL survival (Fig. 5A). However, when an equal number require DC-mediated support for survival and that DCs at both − + ofDCsorCD11cCD11b myeloid cells were cocultured with T- primary and metastatic sites are comparably poised to promote ALL cells, DCs promoted growth of 6.5-fold more viable T-ALL cells survival of circulating T-ALL cells. on average (Fig. 5B). CD3 and CD8 expression confirmed that the majority of viable cells were T-ALL cells and not stromal subsets in PDGFRβ and IGF1R Are Up-Regulated on T-ALL Cells with Concomitant all coculture conditions (Fig. S4D). Collectively, these results dem- Expression of Cognate Ligands by Tumor-Associated DCs. To identify onstrate that myeloid cells, and particularly DCs from the thymic potential pathways governing DC-mediated lymphoma survival, tumor microenvironment, are necessary and sufficient to support we analyzed gene expression profiles of sorted murine T-ALL T-ALL survival and growth. cells for overexpression of growth factor receptors whose ligands were expressed by tumor-associated DCs. Pdgfrb and Igf1r were Thymic DCs Are Functionally Altered in T-ALL. The unique ability of highly expressed by T-ALL cells (Fig. 7A), whereas the ligands tumor-derived stroma to support T-ALL growth ex vivo (Fig. 1 C–E) Pdgfa, Pdgfb, Pdgfc, and Igf1 were expressed by tumor-associated could reflect functional changes in DCs or an increase in DC cel- DCs (Fig. 7B). Expression of these ligands was specific, because + lularity during tumor progression. Thus, we compared the ability of Pdgfd and Igf2 were not expressed. Tumor-associated Sirpα − an equivalent number of WT versus tumor-associated thymic DCs to DCs expressed higher levels of Pdgfc and Igf1 than Sirpα DCs. support T-ALL survival. Only tumor-associated DCs supported Expression data from tumor-associated and WT thymic DC significant T-ALL viability, demonstrating that DCs in the tumor subsets, T-ALL cells, and WT thymocyte subsets are provided in microenvironment have enhanced tumor-promoting capabilities a searchable model within Gene Expression Commons (https:// (Fig. 5C; DC purity shown in Fig. S5A). Distinguishing DCs from gexc.stanford.edu/models/1118/genes) (32). macrophages based on CD11c expression is not always dependable, Immunostaining and flow cytometry confirmed high expression especially in tumor microenvironments (29). LN3 tumor-associated of PDGFRβ proteinonT-ALLcellsinthymicandmetastatictu- DCs are phenotypically consistent with assignment to the DC lineage mors from LN3 and LMO2 mice (Fig. 7C and Fig. S6 A and B), in that they express high levels of CD11c, MHC-II, and CD80 (Fig. while PDGFRβ is expressed mainly on in WT thymi. In- S2). To further confirm the DC identity of tumor-supportive cells, terestingly, PDGFRβ expression is apparent on sparse thymocytes in we transcriptionally profiled DCs from LN3 thymic lymphomas and some pretumor tissues (Fig. 7D and Fig. S6C), suggesting WT thymi, revealing that tumor-associated DCs express the DC- PDGFRβ up-regulation is a characteristic of T-ALL development specific transcription factor Zbtb46 (30) (https://gexc.stanford.edu/ rather than a general peculiarity of thymocytes in LN3 and models/1118/genes). Furthermore, hierarchical clustering of tran- LMO2 murine T-ALL models. MEDICAL SCIENCES scriptional profiling data from tumor-associated DCs, WT thymic We also evaluated PDGFRβ expression in T-ALL patient thymic DCs, and other WT myeloid subsets revealed that tumor-associated and metastatic samples. One of eight patient lymph DCs were most closely related to their WT thymic DC counterparts nodes evaluated displayed elevated PDGFRβ expression on tumor and were distinct from macrophages and monocytes (Fig. 5D)(31). cells (Fig. 7E), indicating that PDGFRβ overexpression occurs in a subset of T-ALL patients. Analysis of previously reported expres- DCs from Metastatic Tumors Support T-ALL Growth ex Vivo. As in the sion profiling data from primary patient samples revealed higher primary thymic tumor, T-ALL cells from metastatic lymph nodes expression of PDGFRB in the early T-cell precursor (ETP) T-ALL and spleens died when cultured alone but survived in the pres- subset (Fig. 7F) (25, 33), suggesting elevated PDGFRβ expression is

Triplett et al. PNAS | Published online February 9, 2016 | E1019 Downloaded by guest on September 27, 2021 ex vivo. Phosphorylation of PDGFRβ (pPDGFRβ) and IGF1R (pIGF1R) was elevated in T-ALL cells relative to WT thymo- cytes (Fig. 8A). We next evaluated whether tumor-associated and/or WT DCs could activate PDGFRβ and IGF1R in T-ALL cells. Tumor-associated DCs sustained activation of IGF1R in lymphoma cells throughout 7 d of culture, whereas WT DCs did not (Fig. 8 B and C). After only 24 h of culture, a time point that precedes of T-ALL cells cultured alone, the majority of T-ALL cells cocultured with tumor-associated DCs had acti- vated IGF1R (Fig. 8B), indicating that tumor-associated DCs signal to the majority of T-ALL cells, rather than to a rare tu- mor-initiating subset. The specificity of pIGF1R and pPDGFRβ staining on T-ALL cells was verified using fluorescence minus one (FMO) and isotype controls, along with specific IGF1R in- hibitors (Fig. S7 A–C). The ability of tumor-associated, but not WT, DCs to sustain IGF1R activation is consistent with higher + gene expression of Igf1 by tumor-associated Sirpα DCs relative to WT DCs (Fig. 8D). Furthermore, tumor-associated DCs secrete elevated levels of IGF1 relative to WT DCs (Fig. 8E). These findings suggest that up-regulation of IGF1 by tumor- associated DCs could enable them to support T-ALL survival. In contrast, PDGFRβ was not activated by tumor-associated DCs (Fig. 8B), despite high expression of the Pdgfc transcript by tumor-associated DCs and activation of PDGFRβ in ex vivo T-ALL cells (Figs. 7A and 8A). Because PDGF-C is latent and thought to inhibit PDGFR signaling until proteolytically acti- vated in the extracellular compartment (36), PDGF-C made by DCs in culture may not be activated to promote PDGFRβ sig- naling in T-ALL cells.

Fig. 5. Tumor-associated thymic myeloid cells are necessary and sufficient to support T-ALL survival. (A) LN3 T-ALL cells were cultured with or without sorted stromal subsets from thymic tumors; viability was assessed after 6 d. Results are normalized to the percentage of live cells in cocultures con- taining All Stroma for each experiment. Graphs depict cumulative data from multiple experiments, with each dot representing the mean of duplicate wells; bars represent means of all experiments. For statistical analysis, the depleted subsets (Left) were compared with All Stroma and enriched subsets (Right) were compared with “No Stroma.” (B) The number of live cells from cocultures with the indicated sorted myeloid subsets in A relative to All Stroma cocultures. (C) LN3 T-ALL cells were cocultured with or without an equivalent number of DCs enriched from LN3 thymic tumors or WT thymi; viability was assessed at day 6. Graphs depict cumulative data, with each line Fig. 6. T-ALL cells at metastatic sites require stromal support, and expanded representing survival from individual experiments (B and C). (D) Hierarchical DCs from metastatic lymphoid organs are sufficient to support T-ALL survival. clustering of the indicated cell subsets was performed using global gene (A) T-ALL cells from thymi, lymph nodes, or spleens were cultured with or expression data. Red text indicates tumor-associated DC subsets. *P < 0.05, without tumor-associated stroma from the same organ, as in Fig. 1; viability was **P < 0.01, ***P < 0.001, and ****P < 0.0001. assessed at day 6. Each line represents data from an individual experiment. (B) Mean fold change in cellularity of the indicated cell types in lymph nodes and spleens of tumor-bearing LN3 mice relative to WT mice. Graphs depict the associated with a more aggressive form of T-ALL (34). Medyouf means + SD of cumulative data (n = 5, LN3; n = 2, WT). The dotted line indicates et al. previously reported that IGF1R is a transcriptional target of a fold change of 1. (C) Single-cell suspensions of splenic T-ALL cells were cul- NOTCH1 that is expressed on human T-ALL cells (35). Together, tured alone, following depletion of DCs, or in the presence of additional tumor- β associated splenic DCs; T-ALL viability was assessed after 6 d. Bars represent these findings reveal that elevated expression of PDGFR and + IGF1R are shared characteristics of mouse and human T-ALL. means SD from five experiments. (D) T-ALL cells from thymi, lymph nodes, or spleens, as indicated, were cultured with or without splenic DCs from the same tumor-bearing mice. (E) Splenic T-ALL cells were cultured after myeloid de- Tumor-Associated DCs Activate IGF1R in T-ALL Cells. To determine pletion or with addition of DCs from the or from the same whether the T-ALL tumor microenvironment supplies endoge- tumor-bearing mouse. (D and E) Lines connect the average of duplicate wells nous signals that activate PDGFRβ and IGF1R, we analyzed the for each condition from individual experiments. *P < 0.05, **P < 0.01, ***P < phosphorylation status of these receptors in tumor cells directly 0.001, and ****P < 0.0001.

E1020 | www.pnas.org/cgi/doi/10.1073/pnas.1520245113 Triplett et al. Downloaded by guest on September 27, 2021 these data indicate that IGF1R signaling activates the MAPK PNAS PLUS pathway and promotes proliferation of T-ALL cells.

DC-Mediated T-ALL Survival Is Dependent on IGF1R Signaling. IGF1 has been shown to directly support growth of solid and hema- tological malignancies (35, 39–41). To determine whether IGF1R signaling was required for DC-mediated T-ALL survival, we treated thymic and splenic T-ALL:DC cocultures with graded concentra- tions of three structurally distinct IGF1R inhibitors: AG-1024 (42), picropodophyllin (PPP) (43), and BMS-754807 (44). All three in- hibitors decreased survival of primary and metastatic T-ALL cells in a dose-dependent manner (Fig. 9A), with IC50 values in the range reported for other tumor studies (40, 41, 44–46). AG-1024 and PPP were chosen because they are reported to have few off-target effects (45, 46); we validated that AG-1024 and PPP inhibited IGF1R activation (Fig. S7C). Notably siRNA-mediated knock down of Igf1r also resulted in a reduction in T-ALL cells expressing pIGF1R (Fig. 9B), with a concomitant decrease in T-ALL survival (Fig. 9C) and proliferation (Fig. 9D). These findings demonstrate that DC-mediated T-ALL survival in the LN3 model is IGF1R- dependent. To determine whether DC-mediated survival and activation of IGF1R, ERK, and PDGFRβ are common features of independent models of T-ALL, we evaluated LMO2 transgenic tumors. As in the

Fig. 7. Elevated expression of IGF1R and PDGFRβ on T-ALL cells and ex- pression of ligands by tumor-associated DCs. (A and B) Relative gene ex- pression of the indicated growth factor receptors in T-ALL cells and WT thymocyte subsets (A) and associated growth factors in tumor-associated DC subsets, T-ALL cells, and WT thymocytes (B). (A and B) Expression is nor- malized to WT DP69- thymocytes. Graphs depict the means + SD of data from three experiments. (C) Representative images (n = 3 each) of a WT thymus and an LMO2 and LN3 thymic lymphoma immunostained for PDGFRβ. (Scale bars: 100 μm.) (D) Representative images of cryosections from three pretumor LN3 thymi immunostained for PDGFRβ. (Scale bars: 100 μm.) (E) Representative images of a nonmalignant and metastatic T-ALL human lymph node (LN) immunostained for PDGFRβ. (Scale bars: 50 μm.) Over- expression of PDGFRβ was seen on one LN of eight patient sections (four LN

and four mediastinal mass). (F)Log2 gene expression values for PDGFRB in ETP (n = 12) and non-ETP (n = 66) T-ALL cases. P values were generated by pairwise t tests in Limma (P = 0.0002; false-discover rate, 0.012). *P < 0.05, **P < 0.01, and ***P < 0.001. Fig. 8. Tumor-associated DCs sustain IGF1R activation but not PDGFRβ. (A) Representative flow cytometric plots of pPDGFRβ and pIGF1R in viable Activation of the MAPK Pathway Is Associated with IGF1R Signaling in cells from WT and tumor-bearing LN3 thymi examined ex vivo. Graphs depict T-ALL Cells. Activation of the MAPK pathway has been implicated the mean fluorescence intensity (MFI) fold change of pPDGFRβ and pIGF1R in in pathogenesis of several (37). Compared with WT viable T-ALL cells versus WT thymocytes; each dot represents an individual thymocytes, ex vivo thymic lymphoma cells had elevated MAPK tumor/WT control analyzed in the same experiment; bars represent the ± signaling, as indicated by increased levels of phosphorylated Erk means SD. The dotted line indicates a fold change of 1. (B and C) Repre- sentative flow cytometric plots of pPDGFRβ and pIGF1R levels in viable (pErk) (Fig. S8A). Coculture with tumor-associated, but not WT, thymic LN3 T-ALL cells after 24 h (B)or7d(C) in culture with DCs enriched thymic DCs sustained pErk over 7 d (Fig. S8A). IGF1R signaling from WT or tumor-bearing LN3 thymi. Graphs depict the MFI fold change of

can activate the MAPK cascade, supporting survival of mature T pPDGFRβ and pIGF1R in T-ALL cells cultured with tumor-associated versus WT MEDICAL SCIENCES cells and other tumors (38, 39). Therefore, we evaluated coex- DCs; dots represent individual tumors; bars represent means ± SD. (D) Rel- ative Igf1 expression levels in DC subsets from WT and tumor-bearing LN3 pression of pIGFR1 with pErk. Levels of pIGF1R consistently cor- − related with pERK levels both in ex vivo T-ALL cells and following thymi normalized to LN3 Sirpα DCs. Means + SD from two WT DC and three coculture with tumor-associated DCs (Fig. S8 B–D). Specificity of LN3 DC datasets are shown. (E) Quantification of IGF1 concentrations in the supernatants of wells containing 1 × 105 DCs enriched from WT (n = 2) or pERK staining was verified with FMO and isotype controls (Fig. S7 LN3 tumor-bearing thymi (n = 3) or from thymic T-ALL cells depleted of + + A and B). Furthermore, inhibiting IGF1R signaling resulted in de- CD11c and CD11b myeloid cells (n = 3). Cultures were established for 24 h, creased pErk levels (Fig. S8E). IGF1R activation also correlated and duplicate wells were averaged for each experiment. *P < 0.05, **P < + with T-ALL cellular proliferation (Ki67 )(Fig. S8F). Together, 0.01, and ****P < 0.001.

Triplett et al. PNAS | Published online February 9, 2016 | E1021 Downloaded by guest on September 27, 2021 Fig. 9. IGF1R signaling is required for DC-mediated T-ALL survival. (A) Primary LN3 splenic or thymic T-ALL cells were cultured for 4 d with tumor-associated DCs from the same organ before addition of the IGF1R small-molecule inhibitors AG-1024, PPP, or BMS-754807, as indicated. Viability was assessed af- ter 3 d. Bars represent the means + SD from three to seven experiments for each condition normalized to cocultures containing a DMSO vehicle control.

The IC50 values are indicated above the graphs. (B–D) T-ALL cells were cultured with tumor-associated DCs in the presence or absence of IGF1R-targeted siRNA (2.5 μM) or a nontargeted siRNA control (2.5 μM); 6 d after culture initiation, T-ALL cells were evaluated by flow cytometry for pIGF1R levels (B), viability (C), and IGF1R activation versus pro- liferation by Ki67 (D). Flow cytometric plots are representative of four independent experiments. The graph in B depicts the percentage reduction of + pIGF1R T-ALL cells in the indicated siRNA treatment groups compared with control wells without siRNA. The graph in C depicts T-ALL viability normalized to control wells without siRNA. Graphs in B and C rep- resent cumulative data + SD from four tumors an- alyzed in independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

LN3 model, LMO2 T-ALL cells survived only if cocultured with lymphoid organs prototypically promote T-cell proliferation and tumor-associated DCs (Fig. S9A). When examined directly ex vivo, survival during T-cell activation and in homeostasis. In keeping LMO2 T-ALL cells had elevated levels of pIGF1R, pErk, and with this function in the periphery, a previous study demonstrated pPDGFRβ relative to WT thymocytes (Fig. S9B). Furthermore, that DCs could directly support survival of cutaneous T-cell lym- IGF1R phosphorylation correlated with Erk phosphorylation in phoma cells in vitro (47). Additionally, T-ALL–promoting activity T-ALL cells analyzed ex vivo and correlated with ERK activation was not exclusive to DCs located in primary thymic tumors, and proliferation in T-ALL cells cocultured with tumor-associated because DCs from tumor-bearing spleens and thymi promoted T- DCs (Fig. S9 C and D). Importantly, as in the LN3 model, inhibition ALL survival equivalently. Furthermore, tumor cells at both pri- of IGF1R signaling abrogated DC-mediated survival of LMO2 mary and metastatic sites required the presence of DCs for sur- T-ALL cells (Fig. S9E). Altogether, these studies provide the first vival. Thus, T-ALL cells first interact with altered DCs in the indication to our knowledge that DCs in the tumor microenviron- thymic tumor environment that can promote their survival, and as ment directly support survival and growth of T-ALL cells in an T-ALL cells seed metastatic sites, they encounter peripheral DCs IGF1R-dependent manner. poised to sustain their growth. DCs were also present throughout patient mediastinal mass sections and lymph nodes, indicating that Discussion DCs are positioned to support human T-ALL growth as well. In this study, we demonstrate that the thymic microenvironment Collectively, these results suggest that targeted therapies against becomes altered during LN3 T-ALL progression such that DCs gain DCs or their tumor-promoting signals would affect tumor growth the capacity to support T-ALL growth. Human primary T-ALL at both primary and metastatic tumor sites. samples require stromal-derived signals for survival (10, 16), sug- We found that IGF1R signaling was required for DC-mediated gesting such signals, if identified, could be viable therapeutic targets. murine T-ALL growth in two molecularly distinct murine models of Previous studies revealed that in the presence of exogenous cyto- T-ALL. Exon sequencing revealed that many LN3 tumors have ac- kines, TECs from healthy thymi and bone marrow (BM) cell lines tivating Notch1 mutations, and some LMO2 tumors are also could support T-ALL (10, 16–18). However, endogenous tumor NOTCH-driven (25). Thus, DCs and IGF1R signaling can promote stromal cells capable of supporting T-ALL growth in the absence of survival of NOTCH-driven murine T-ALL cells. All of the LN3 and exogenous cytokines have not been previously identified. LMO2 tumors analyzed required DC-mediated support for survival We initially focused on TECs because they express IL-7 and in vitro, but some of the LN3 tumors did not have detectable acti- NOTCH ligands, which have been implicated in supporting vating mutations in Notch1, suggesting that DCs may also support T-ALL. However, we found that epithelial-free regions enriched growth of other molecular subsets of T-ALL. IGF1R activation is for proliferating lymphoma cells emerged at the periphery of likely relevant to human T-ALL growth as well. Activating mutations thymic tumors in two distinct T-ALL models, consistent with in NOTCH1 have been reported in >50% of T-ALL cases (22), and epithelial-free regions in thymic tumors that arise in AKR mice activated NOTCH1 induces IGF1R expression in mouse and human (23). Epithelial-free regions also comprised the majority of tissue T-ALL cells (35, 48). Importantly, IGF1R expression contributes in mediastinal mass sections from T-ALL patients. These find- to T-ALL leukemia-initiating cell activity in vivo, and inhibition ings, along with the fact that TECs are confined to thymi and, of IGF1R signaling prolongs survival following transplantation of thus, are not present to support T-ALL growth at metastatic T-ALL in mice (35). Previous studies in other cancers have dem- sites, indicated that other stromal subsets must support T-ALL. onstrated that IGF1R signaling confers resistance to traditional Unexpectedly, we found that tumor-associated myeloid cells, therapies (49), providing a rationale for combining IGF1R-targeted particularly DCs, were necessary and sufficient to support survival therapy with standard treatments (e.g., and radiation) and growth of T-ALL cells in vitro. Given that thymic DCs gen- or in conjunction with inhibitors of other pathways. erally induce negative selection of autoreactive thymocytes, it was Although we found that IGF1R signaling was necessary for surprising that these DCs would provide survival signals to tumors DC-mediated T-ALL survival, other myeloid-derived factors derived from developing T cells. In contrast, DCs in secondary likely contribute to T-ALL progression in vivo. Transwell assays

E1022 | www.pnas.org/cgi/doi/10.1073/pnas.1520245113 Triplett et al. Downloaded by guest on September 27, 2021 revealed that T-ALL cells required close contact with tumor In summary, we identify DCs as an endogenous cell type in the PNAS PLUS stroma to survive. Because IGF1 is secreted, other transmembrane tumor microenvironment that becomes altered during T-ALL molecules may therefore mediate interactions between tumor progression to gain the capacity to directly support tumor growth cells and DCs. Tumor–DC interactions could promote local- through IGF1R signaling. To our knowledge, this is the first ized delivery of IGF1 to T-ALL cells and/or could induce sig- demonstration that (i) DCs from the tumor microenvironment naling through alternate mitogenic pathways. In addition, gene directly support T-ALL survival, (ii) DCs are a source of IGF1 ontology (GO) analysis comparing tumor-associated versus WT mitogenic signals that promote T-ALL survival, and (iii) PDGFRβ thymic DCs revealed that tumor-associated DCs have increased is up-regulated on an aggressive subset of human T-ALLs and expression of genes involved in “response to ” (P = is activated by the tumor microenvironment in two distinct murine − − 21.4 × 10 11) and “inflammatory responses” (P = 1.3 × 10 8). T-ALL models. Notably, DCs can prime antitumor adaptive im- Some of these genes, which are associated with alternatively mune responses; thus, different DC subsets likely have the op- activated (M2) macrophages and TAMs (Chi3l3, Retnla,and posing roles of either promoting T-ALL growth, as shown here, or Arg1), have been implicated in growth of several tumors (4, 50, controlling T-ALL growth via activation of antitumor immune re- 51). Like Igf1, these genes were up-regulated to a greater degree + sponses. It will be important to determine which specific DC in the Sirpα subset of thymic DCs, which are numerically ex- subsets directly support T-ALL growth, so as to assess the panded in thymic lymphomas (https://gexc.stanford.edu/models/ + therapeutic potential of targeting these subsets and their tumor 1118/genes). These findings suggest that Sirpα DCs are most promoting signals in vivo, while leaving intact the capacity to capable of supporting T-ALL growth, although our data dem- mount antitumor immune responses. Thus, our findings pro- onstrate that other myeloid cells likely contribute. Similarly, vide a strong rationale for further investigation of DC-mediated monocytes differentiated into M2 macrophages were shown to T-ALL growth to identify novel therapeutic targets. directly promote proliferation of adult T-cell leukemia/lymphoma cell lines (52). Thus, it will be important to evaluate the respective Materials and Methods + − roles of tumor-associated Sirpα and Sirpα DCs, as well as other Mice. LN3 mice were provided by Tom Serwold, Joslin Diabetes Center, Boston myeloid subsets in supporting T-ALL growth. (21). C57BL/6J and CD11c-EYFP [B6.Cg-Tg(Itgax-EYFP)1Mnz/J] mice were It is not clear why Pdgf genes expressed by tumor-associated DCs purchased from The Jackson Laboratory. LN3 and LMO2 (B6.CD2-Lmo2) (25) didnotsustainactivationofPDGFRβ in T-ALL cells in vitro. Al- mice were euthanized upon lymphoma detection, as identified by enlarged though Pdgfa and Pdgfb may not be expressed at sufficiently high lymph nodes, hunched posture, reduced motility, ruffled fur, and/or labored + levels, Pdgfc was expressed robustly by tumor-associated Sirpα breathing. All mice were housed under specific pathogen-free conditions at The University of Texas at Austin (UT Austin) and the Eunice Kennedy Shriver DCs; however, PDGF-C stimulates PDGFRαα and αβ hetero- ββ National Institute of Child Health and Human Development, National Insti- dimers, instead of the PDGFR receptor found to be up-regulated tutes of Health (NICHD/NIH). Experimental procedures were approved by the on LN3 T-ALL cells (36). Furthermore, PDGF-C and -D block Institutional Animal Care and Use Committee at UT Austin or the Animal binding to PDGFRββ before their activation by extracellular pro- Care and Use Committee at the NICHD/NIH. teolytic cleavage, which may not occur in in vitro DC cocultures (36). Although tumor-associated DCs were not sufficient to sustain acti- Notch1 Exon Sequencing. DNA was extracted from 5 × 106 WT thymocytes or vation of PDGFRβ in cocultures, the receptor was phosphorylated primary LN3 T-ALL cells using the DNeasy Blood and Tissue Kit (Qiagen). Notch1 in T-ALL cells analyzed ex vivo, indicating the tumor microenvi- exons26and27wereamplifiedwiththefollowingprimersets:26forward, ronment activates PDGFRβ. Extensive vascularization and fibro- 5′-ACGGGAGGACCTAACCAAAC-3′;26reverse,5′-CAGCTTGGTCTCCAACACCT-3′; blast/ECM networks were present in the aberrant T-ALL tumor 27 forward, 5′-CGCTGAGTGCTAAACACTGG-3′; and 27 reverse, 5′-GTTTTGCCTG- ′ microenvironment, and fibroblasts and endothelial cells may pro- CATGTACGTC-3 .Exon34wasamplifiedintwofragmentsusingthefollowing primer sets: 34.1 forward, 5′-GCTCCCTCATGTACCTCCTG-3′; 34.1 reverse, 5′-TAGTG- vide high levels of PDGFs and activating proteases in vivo (36). ′ ′ ′ β GCCCCATCATGCTAT-3 ;34.2forward,5-ATAGCATGATGGGGCCACTA-3 ;and Thus, PDGFR signaling may be sustained by DCs and/or other 34.2 reverse, 5′-CTTCACCCTGACCAGGAAAA-3′. These primer sets have been tumor stromal subsets in the tumor microenvironment, which may previously described (57). PCR products were purified using a PCR purification contribute to T-ALL progression. (Qiagen) and sequenced directly at the UT Austin Institute for Cellular and We hypothesize that PDGFRβ is also relevant to human Molecular Biology Genome Sequencing and Analysis Facility. Notch1 sequences T-ALL, because PDGFRB expression was elevated in a subset from T-ALL were compared against those from WT thymocytes using the of T-ALL patient samples, and PDGFRβ was overexpressed on novoSNP analysis program (58). A cutoff score of 1 and F score of 1 were used T-ALL cells in a primary patient lymph node. Furthermore, to identify mutations that were then verified only if forward and reverse PDGF receptors are up-regulated in mature T-cell lymphomas primers revealed the same . The ApE program (A plasmid Editor (53–56), and a translocation involving PDGFRB has been iden- version 2.0.49 by M. Wayne Davis, University of Utah, Salt Lake City) was used tified in a T-ALL patient (53). To our knowledge, we provide for translating exon-sequencing results into protein sequences. β the first evidence that PDGFR is up-regulated preferentially Patient Samples. Lymph node and mediastinal mass tissue samples from T-ALL in ETP T-ALLs. Gene expression profiling of WT thymocytes patients were collected and provided by the Texas Children’s Hospital, and β demonstrates that PDGR is normally expressed on ETPs and deidentified, formalin-fixed, paraffin-embedded tissue sections were pro- subsequently down-regulated as the cells differentiate (https:// vided for analysis with approval from the institutional review board com- gexc.stanford.edu/models/1118/genes). Thus, inappropriate main- mittees at UT Austin and Texas Children’s Hospital. tenance of this may contribute to tumorigenesis of the immature and aggressive ETP T-ALL subset. Importantly, Immunostaining and Microscopy. Immunostaining of murine tissue was per- remission of a patient with refractory anaplastic large cell lym- formed on 7-μm acetone-fixed cryosections according to standard protocols. – phoma was recently achieved with a PDGFRβ inhibitor, The following antibodies were used: anti-CD11c (N418), anti Pan-Endothelial MEDICAL SCIENCES β Cell (MECA-32), anti-fibroblast/ECM (ER-TR7), anti-Keratin5 (poly- (55), highlighting the potential of targeting PDGFR in other α – T-cell malignancies. However, our studies demonstrate that DCs clonal; Covance), anti-Sirp (p84), anti Pan-Cytokeratin (C-11), anti-Keratin8 (TROMA-1), anti-CD3 (17A2), and anti-PDGFRβ (APB5). To detect proliferation, promote T-ALL survival through IGF1R activation in the ab- ′ β β mice were injected intraperitoneally with 2 mg of 5-bromo-2 -deoxyuridine sence of PDGFR signaling. Thus, whereas targeting PDGFR (BrdU) 20 min before euthanasia, and thymic cryosections were stained with may be warranted in a subset of T-ALL patients with up-regu- anti-BrdU and anti-CD8 (53-6.7) per standard protocols. lated receptors, PDGFRβ inhibition may be efficacious only in Immunostaining of formalin-fixed, paraffin-embedded patient tissue samples combination with other standard regimens or targeted agents, followed standard protocols. Slides were incubated with: anti–Pan-Cytokeratin such as IGF1R inhibitors. (AE1/AE3), anti-CD11c (EP1347Y), and anti-PDGFRβ (28E1). Primary antibodies

Triplett et al. PNAS | Published online February 9, 2016 | E1023 Downloaded by guest on September 27, 2021 were detected with fluorochrome-conjugated anti-Ig (Jackson ImmunoResearch), Inhibitor and siRNA Studies. IGF1R inhibitors picropodophyllin (EMD Milli- or streptavidin (Life Technologies). pore), tyrphostin AG-1024, and BMS-754807 (Selleckchem) were resuspended Images were acquired using an Axio Imager A1 microscope (Zeiss), with an in DMSO according to the manufacturer’s protocols for use in cell cultures. AxioCam MRm camera (Zeiss); Zeiss EC Plan-Neofluar 5×/0.16, Plan-Apochromat Equal volumes of DMSO were used as vehicle controls. For IGF1R siRNA 6 5 10×/0.45, and Plan-Apochromat 20×/0.8 objectives; and AxioVision 4.8 software knockdown experiments, 2 × 10 T-ALL cells with 1 × 10 DCs were cultured (Zeiss). Adobe Photoshop version 12.0 was used for overlays of fluorescent for 24 h in 0.4 mL and then treated with 2.5 μM Accell IGF1R-targeted channels and to adjust exposure uniformly across the images. Composite SMARTpool siRNA or Accell Nontargeting Pool siRNA for an additional 6 d immunofluorescence images were acquired with an Eclipse 80i microscope according to the manufacturer’s protocol (Dharmacon). (Nikon), a CoolSNAP HQ2 camera (Photometrics), and a 10×/0.3 CFI Plan Fluor objective (Nikon) and were stitched using NIS Elements 3.0 (Nikon). Two-Photon Fluorescence Microscopy. Two-photon imaging of thymocytes or tumor cells migrating within thymic slices was carried out as described pre- μ Flow Cytometry. For thymic stromal subsets, WT and tumor-bearing thymi were viously (27). Briefly, 400- m thymic slices from CD11c-EYFP transgenic mice enzymatically digested with a combination of Collagenase IV and Dispase as (28) were prepared using a vibratome (Leica), and WT thymocytes or LN3 described (59). Single-cell suspensions were immunostained with Ulex euro- tumor cells, labeled with CellTracker CMTPX Red or CMF2HC Blue (Invi- – paeus agglutinin I (UEA-1), anti-CD3 (145-2C11), anti-CD8 (53-6.7), anti-CD11c trogen), were incubated with thymic slices for 1 4 h at 37 °C in 5% CO2. μ μ (N418), anti-I-A/I-E (M5/114.15.2), anti-CD45 (30-F11), anti-CD11b (M1/70), anti- Images were acquired every 15 s, through a depth of 40 m, in 5- m intervals B220 (RA3-6B2), anti-TER-119 (TER-119), anti-CD31 (390), anti-EpCAM (G8.8), with an Ultima IV microscope (Prairie) equipped with two MaiTai lasers anti-Sirpα (P84), anti-Ly-51 (6C3), anti-CD4 (GK1.5), anti-CD80 (16-10A1), or anti- (Spectra-Physics) and PrairieView acquisition software. Excitation wave- PDGFRβ (APB5). Intracellular phosphorylated protein levels were evaluated us- lengths of 840 nm, or 740 and 900 nm for three-color imaging, were used; ing the FIX & PERM (fixation and permeabilization) kit (Invitrogen), according 480/40, 535/50, and 607/45 bandpass filters (Chroma Technology) were used to the manufacturer’s methanol-modified protocol, and stained with anti- for detection of CMF2HC, EYFP, and CMTPX fluorescence, respectively. pPDGFRβ (J24-618), anti-pIGF1R (K74-218), or anti-pErk1/2 (20A). Equal Imaris software version 7.6 (Bitplane) was used for cell tracking. Cellular concentrations of mouse IgG1 isotype antibodies (MOPC-21) were used as interactions were scored when the distance between cell surfaces was less than 3 μm. The percentage of time contacting DCs equals the percentage of all negative controls for intracellular staining. Viability for in vitro studies was tracked time points in which a cell contacts a DC. Dwell time equals the total assessed by flow cytometry with Annexin-V and propidium iodide (PI), and time during which thymocytes contact DCs in consecutive time points. DNA content was assessed using DyeCycle Violet (Invitrogen) according to manufacturer’s protocols. Viable lymphoma cells were identified with anti- TCRβ (H57-957), anti-CD8a (53-6.7), anti-CD4 (RM4-5), and anti-CD3e (145- Global Gene Expression Analysis. Genome-wide expression analysis was 2C11). An LSRFortessa or a FACSAria II (BD Biosciences) was used for analysis performed on T-ALL cells and DC subsets sorted from LN3 thymic tumors using Affymetrix GeneChip Mouse Genome 430 2.0 Arrays (n = 3, LN3 and sorting, and data were analyzed with FlowJo (Treestar). + − tumor Sirpα DCs and Sirpα DCs; n = 3, T-ALL cells). We have previously reported expression profiling of WT thymic DCs and thymocytes (32, 60) Cell Culture Media. Complete cell culture media consisted of Roswell Park [Gene Expression Omnibus (GEO) accession nos. GSE34723 and GSE56928]; Memorial Institute 1640 medium (RMPI 1640) supplemented with 1× non- + − 1 μg of total RNA from 3 × 104 sorted tumor-associated Sirpα DCs, Sirpα essential amino acids, 55 μM β-mercaptoethanol, 2 mM GlutaMAX, 1× DCs, or T-ALL cells was amplified, labeled, and hybridized according to penicillin–streptomycin–glutamine (100 U/mL penicillin, 100 μg/mL strep- the manufacturer’s specifications. Data were uploaded to Gene Expres- tomycin, 292 μg/mL glutamine), 1 mM sodium pyruvate (all from Gibco), and sion Commons (32) for normalization and analysis; an in silico model 10% (vol/vol) FBS (Thermo Scientific HyClone). All cultures were maintained containing these datasets, along with WT thymocytes and DCs from at 37 °C in 5% CO . 2 C57BL/6J mice at 6 mo of age, is available in Gene Expression Commons (https://gexc.stanford.edu/models/1118/genes);thedatahavealsobeen Tumor: Stromal Cocultures. Cocultures of ex vivo lymphoma cells with bulk deposited in GEO (GSE54609). For hierarchical clustering (Euclidean stroma isolated from tumors or WT thymi were prepared and cultured as complete with average linkage), we used R package Easy Microarray data previously described (20). Briefly, lymphoma cells were isolated by strain- Analysis(EMA)(61)toclusterthymic 6-mo WT or tumor-associated LN3 + − ing either thymic, lymph node, or splenic tumor-bearing tissue through a DC subsets, WT CD4 CD8 thymocytes, and the following peripheral μ + − 40- m filter. Stromal cells were acquired by mincing tumor-bearing or WT myeloid subsets: common myeloid progenitors (CMP), CD8 DCs, CD8 tissue in complete RPMI on ice. Fragments were then suspended in 10 mL of DCs, plasmacytoid DCs (pDCs), monocytes, BM macrophages (BM Mac.), complete RPMI and vortexed. After aggregates settled for 3 min, 7 mL of and peritoneal macrophages (Mac.). Reference datasets were downloaded supernatant was replaced with fresh RPMI to deplete lymphoma cells in from GEO (accession nos. GSE10246 and GSE34723) (31, 32) and normalized solution, and the aggregates were vortexed. This process was repeated a with thymic DC datasets using GC robust multi-array average (GCRMA). GO 5 total of three times before remaining cells were counted; 5 × 10 lym- analyses were carried out using the database for annotation, visualization, and 5 phoma cells were then cultured alone, or 2.5 × 10 lymphoma cells were integrated discovery (DAVID) Bioinformatics Database (62). 5 5 cultured with 2.5 × 10 bulk stromal cells. For transwell assays, 5 × 10 Human T-ALL gene expression datasets were generated with Affymetrix lymphoma cells were seeded in the upper chamber of a 0.4-μm transwell U133A chips and comprised ETP and non-ETP cases. CEL files were kindly 5 insert dish (Corning Incorporated Costar), and 5 × 10 cells were seeded in provided by Charles Mullighan from cases at St. Jude Children’s Research the lower chambers; when both lymphoma cells and thymic tumor stromal Hospital, Memphis, TN, and have been previously described (GEO accession cells were present in the lower chambers, they were mixed at a 1:1 ratio no. GSE33315) (25, 33). Microarray data were normalized using the Robust with 2.5 × 105 cells each. MultiChip Averaging (RMA) algorithm (58), as implemented in the Bio- For cocultures of tumor cells with FACS-sorted thymic stromal subsets, 2 × 106 conductor package Affy. For pairwise group comparisons, we used t test in T-ALL cells were cultured with sorted stromal subsets (5 × 105 All Stroma, 5 × 105 the Limma package (59) in Bioconductor to identify differentially expressed depleted populations, 2.5 × 105 sorted myeloid, 1 × 105 sorted DCs, or 1 × 105 probe sets between ETP and non-ETP cases under comparison. The imple- sorted CD11c−CD11b+ cells) in 1 mL of RPMI in 48-well plates. mentation of t test in Limma uses an empirical Bayes method to moderate Enriched DCs from tumors and WT tissue were isolated from tissue using the SEs of the estimated log-fold changes, and P values were corrected using biotinylated anti-CD11c (N418) and streptavidin microbeads with MACS the false-discovery method. columns (Miltenyi Biotech) according to the manufacturer’s instructions. Myeloid cells were further depleted using biotinylated anti-CD11b (M1/70), Statistical Analysis. Statistical significance was determined with one sample, as indicated; 2 × 106 lymphoma cells were cultured alone or with 1 × 105 DCs. paired and unpaired Student’s t tests using GraphPad Prism software (GraphPad). DC purities were determined by flow cytometry using anti-biotin (Bio3- 18E7), anti-Sirpα (P84), and anti-I-A/I-E (M5/114.15.2). ACKNOWLEDGMENTS. We thank Coby Tran for assistance with immuno- staining, Ellen R. Richie for helpful discussions, and Michele Redell for her advice IGF1 ELISA. DCs were enriched from WT thymi or LN3 thymic lymphomas, as and expertise in human hematologic malignancies. We also thank the staff at × 5 The University of Texas at Austin Animal Facilities. This work was supported by described in Tumor: Stromal Cocultures;1 10 DCs or T-ALL cells were then Recruitment Award R1003 from the Cancer Prevention and Research Institute of μ plated in 250 L in 48-well plates. Supernatants were collected after 24 h and Texas (to L.I.R.E.) and by the Intramural Research Program of the NIH (PEL: snap frozen at −80 °C. IGF1 levels were measured using a sandwich ELISA Project number: 1ZIAHD001803-22). T.A.T. was supported by Postdoctoral according to the manufacturer’s protocol (RayBiotech). Fellowship PF-14-216-01-TBG from the American Cancer Society.

E1024 | www.pnas.org/cgi/doi/10.1073/pnas.1520245113 Triplett et al. Downloaded by guest on September 27, 2021 1. Van Vlierberghe P, Ferrando A (2012) The molecular basis of acute lympho- 31. Lattin JE, et al. (2008) Expression analysis of G Protein-Coupled Receptors in mouse PNAS PLUS blastic leukemia. J Clin Invest 122(10):3398–3406. macrophages. Immunome Res 4(1):5. 2. Egeblad M, Nakasone ES, Werb Z (2010) Tumors as organs: Complex tissues that in- 32. Seita J, et al. (2012) Gene Expression Commons: An open platform for absolute gene terface with the entire organism. Dev Cell 18(6):884–901. expression profiling. PLoS One 7(7):e40321. 3. Bhowmick NA, Neilson EG, Moses HL (2004) Stromal fibroblasts in cancer initiation 33. Zhang J, et al. (2012) The genetic basis of early T-cell precursor acute lymphoblastic and progression. Nature 432(7015):332–337. leukaemia. Nature 481(7380):157–163. 4. Lawrence T, Natoli G (2011) Transcriptional regulation of macrophage polarization: 34. Coustan-Smith E, et al. (2009) Early T-cell precursor leukaemia: A subtype of very high- Enabling diversity with identity. Nat Rev Immunol 11(11):750–761. risk acute lymphoblastic leukaemia. Lancet Oncol 10(2):147–156. 5. Lewis CE, Leek R, Harris A, McGee JO (1995) Cytokine regulation of angiogenesis in 35. Medyouf H, et al. (2011) High-level IGF1R expression is required for leukemia-initi- breast cancer: The role of tumor-associated macrophages. J Leukoc Biol 57(5): ating cell activity in T-ALL and is supported by Notch signaling. J Exp Med 208(9): 747–751. 1809–1822. 6. Cieslewicz M, et al. (2013) Targeted delivery of proapoptotic peptides to tumor- 36. Cao Y (2013) Multifarious functions of PDGFs and PDGFRs in tumor growth and associated macrophages improves survival. Proc Natl Acad Sci USA 110(40):15919–15924. metastasis. Trends Mol Med 19(8):460–473. 7. Zeisberger SM, et al. (2006) Clodronate--mediated depletion of tumour- 37. Sebolt-Leopold JS, Herrera R (2004) Targeting the mitogen-activated associated macrophages: A new and highly effective antiangiogenic therapy approach. cascade to treat cancer. Nat Rev Cancer 4(12):937–947. BrJCancer95(3):272–281. 38. Brocardo MG, et al. (2001) Early effects of insulin-like growth factor-1 in activated 8. Kimura YN, et al. (2007) Inflammatory stimuli from macrophages and cancer cells human T lymphocytes. J Leukoc Biol 70(2):297–305. synergistically promote tumor growth and angiogenesis. Cancer Sci 98(12):2009–2018. 39. Pollak MN, Schernhammer ES, Hankinson SE (2004) Insulin-like growth factors and 9. Gabrilovich D (2004) Mechanisms and functional significance of tumour-induced neoplasia. Nat Rev Cancer 4(7):505–518. dendritic-cell defects. Nat Rev Immunol 4(12):941–952. 40. Yaktapour N, et al. (2013) Insulin-like growth factor-1 receptor (IGF1R) as a novel 10. Armstrong F, et al. (2009) NOTCH is a key regulator of human T-cell acute leukemia target in chronic lymphocytic leukemia. Blood 122(9):1621–1633. initiating cell activity. Blood 113(8):1730–1740. 41. Shi P, et al. (2009) IGF-IR interacts with NPM-ALK oncogene to induce 11. Love PE, Bhandoola A (2011) Signal integration and crosstalk during thymocyte mi- survival of T-cell ALK+ anaplastic large-cell lymphoma cells. Blood 114(2):360–370. gration and emigration. Nat Rev Immunol 11(7):469–477. 42. Párrizas M, Gazit A, Levitzki A, Wertheimer E, LeRoith D (1997) Specific inhibition of 12. Hu Z, Lancaster JN, Ehrlich LIR (2015) The Contribution of and Migration insulin-like growth factor-1 and tyrosine kinase activity and biological to the Induction of Central Tolerance in the Thymus. Front Immunol 6:398. function by tyrphostins. Endocrinology 138(4):1427–1433. 13. Shitara S, et al. (2013) IL-7 produced by thymic epithelial cells plays a major role in the 43. Girnita A, et al. (2004) Cyclolignans as inhibitors of the insulin-like growth factor-1 development of thymocytes and TCRγδ+ intraepithelial lymphocytes. J Immunol receptor and malignant cell growth. Cancer Res 64(1):236–242. 190(12):6173–6179. 44. Carboni JM, et al. (2009) BMS-754807, a small molecule inhibitor of insulin-like 14. Koch U, et al. (2008) Delta-like 4 is the essential, nonredundant ligand for Notch1 growth factor-1R/IR. Mol Cancer Ther 8(12):3341–3349. during thymic T cell lineage commitment. J Exp Med 205(11):2515–2523. 45. Vasilcanu D, et al. (2004) The cyclolignan PPP induces activation loop-specific in- 15. Hozumi K, et al. (2008) Delta-like 4 is indispensable in thymic environment specific for hibition of tyrosine phosphorylation of the insulin-like growth factor-1 receptor. Link T cell development. J Exp Med 205(11):2507–2513. to the phosphatidyl inositol-3 kinase/Akt apoptotic pathway. Oncogene 23(47): 16. Silva A, et al. (2011) IL-7 contributes to the progression of human T-cell acute lym- 7854–7862. phoblastic . Cancer Res 71(14):4780–4789. 46. Deutsch E, et al. (2004) Tyrosine kinase inhibitor AG1024 exerts antileukaemic effects 17. Scupoli MT, et al. (2007) 7 requirement for survival of T-cell acute lym- on STI571-resistant Bcr-Abl expressing cells and decreases AKT phosphorylation. Br J phoblastic leukemia and human thymocytes on bone marrow stroma. Haematologica Cancer 91(9):1735–1741. 92(2):264–266. 47. Berger CL, et al. (2002) The growth of cutaneous T-cell lymphoma is stimulated by 18. Scupoli MT, et al. (2003) Thymic epithelial cells promote survival of human T-cell acute immature dendritic cells. Blood 99(8):2929–2939. lymphoblastic leukemia blasts: The role of interleukin-7. Haematologica 88(11): 48. Trimarchi T, et al. (2014) Genome-wide mapping and characterization of Notch-reg- 1229–1237. ulated long noncoding RNAs in acute leukemia. Cell 158(3):593–606. 19. Gachet S, et al. (2013) Leukemia-initiating cell activity requires calcineurin in T-cell 49. Weroha SJ, Haluska P (2008) IGF-1 receptor inhibitors in clinical trials–early lessons. acute lymphoblastic leukemia. Leukemia 27(12):2289–2300. J Mammary Gland Biol Neoplasia 13(4):471–483. 20. Richie ER, Walker DA (2000) Production and characterization of immature murine 50. Loke P, et al. (2002) IL-4 dependent alternatively-activated macrophages have a dis- T-lymphoma cell lines. Methods Mol Biol 134:177–184. tinctive in vivo gene expression phenotype. BMC Immunol 3:7. 21. Serwold T, et al. (2010) T-cell receptor-driven lymphomagenesis in mice derived from 51. Raes G, et al. (2002) Differential expression of FIZZ1 and Ym1 in alternatively versus a reprogrammed T cell. Proc Natl Acad Sci USA 107(44):18939–18943. classically activated macrophages. J Leukoc Biol 71(4):597–602. 22. Weng AP, et al. (2004) Activating mutations of NOTCH1 in human T cell acute lym- 52. Komohara Y, et al. (2013) Clinical significance of CD163⁺ tumor-associated macro- phoblastic leukemia. Science 306(5694):269–271. phages in patients with adult T-cell leukemia/lymphoma. Cancer Sci 104(7):945–951. 23. Davey GM, Tucek-Szabo CL, Boyd RL (1996) Characterization of the AKR thymic mi- 53. Piccaluga PP, et al. (2005) Expression of platelet-derived growth factor receptor alpha croenvironment and its influence on thymocyte differentiation and lymphoma de- in peripheral T-cell lymphoma not otherwise specified. Lancet Oncol 6(6):440. velopment. Leuk Res 20(10):853–866. 54. Terada T (2012) TDT (-), KIT (+), CD34 (+), CD99 (+) precursor T lymphoblastic leu- 24. Li J, Park J, Foss D, Goldschneider I (2009) Thymus-homing peripheral dendritic cells kemia/lymphoma. Int J Clin Exp Pathol 5(2):167–170. constitute two of the three major subsets of dendritic cells in the steady-state thymus. 55. Laimer D, et al. (2012) PDGFR blockade is a rational and effective therapy for NPM- J Exp Med 206(3):607–622. ALK-driven lymphomas. Nat Med 18(11):1699–1704. 25. Smith S, et al. (2014) LIM domain only-2 (LMO2) induces T-cell leukemia by two dis- 56. Huang Y, et al. (2010) Gene expression profiling identifies emerging oncogenic path- tinct pathways. PLoS One 9(1):e85883. ways operating in extranodal NK/T-cell lymphoma, nasal type. Blood 115(6):1226–1237. 26. Lutzny G, et al. (2013) Protein kinase c-β-dependent activation of NF-κB in stromal 57. O’Neil J, et al. (2006) Activating Notch1 mutations in mouse models of T-ALL. Blood cells is indispensable for the survival of chronic lymphocytic leukemia B cells in vivo. 107(2):781–785. Cancer Cell 23(1):77–92. 58. Weckx S, et al. (2005) novoSNP, a novel computational tool for sequence variation 27. Ehrlich LIR, Oh DY, Weissman IL, Lewis RS (2009) Differential contribution of che- discovery. Genome Res 15(3):436–442. motaxis and substrate restriction to segregation of immature and mature thymocytes. 59. Gray DHD, et al. (2008) Unbiased analysis, enrichment and purification of thymic Immunity 31(6):986–998. stromal cells. J Immunol Methods 329(1-2):56–66. 28. Lindquist RL, et al. (2004) Visualizing dendritic cell networks in vivo. Nat Immunol 60. Ki S, et al. (2014) Global transcriptional profiling reveals distinct functions of thymic 5(12):1243–1250. stromal subsets and age-related changes during thymic involution. Cell Reports 9(1): 29. van de Ven R, Lindenberg JJ, Oosterhoff D, de Gruijl TD (2013) Dendritic Cell Plasticity 402–415. in Tumor-Conditioned Skin: CD14(+) Cells at the Cross-Roads of Immune Activation 61. Servant N, et al. (2010) EMA - A R package for Easy Microarray data analysis. BMC Res and Suppression. Front Immunol 4:403. Notes 3:277. 30. Meredith MM, et al. (2012) Expression of the zinc finger transcription factor zDC 62. Dennis G, Jr, et al. (2003) DAVID: Database for Annotation, Visualization, and In- (Zbtb46, Btbd4) defines the classical dendritic cell lineage. J Exp Med 209(6):1153–1165. tegrated Discovery. Genome Biol 4(5):P3. MEDICAL SCIENCES

Triplett et al. PNAS | Published online February 9, 2016 | E1025 Downloaded by guest on September 27, 2021